Does BELD5 need water fix?

We are using EPA’s 1-km BELD5 dataset and the accompany tool to generate beld5 file for our own grid.
However, during the verification, we found differences in b3grd emissions on the 12US2 grid between the EPA released b3grd file from the 2017paltform (by windowing out from the 12US1 file) and the one we re-generated from using the beld5 file we created ourselves as input to BEIS3.

One example of such differences can be seen from the plot here:

It seems to me that the difference are all at coastal or water related cells.

I wonder if we were missing a step of water fix here to prepare the beld5 file, like preparing the beld4.1 files?

Thanks in advance.

Yongtao Hu

Georgia Tech


We did make some adjustments for Canadian lakes/water from what I can recall. Where are you getting the 1km BELD5 data specifically from in your processing?
Thanks, Jeff

Hi Jeff,

I used the nc4 version of the 1km BELD5 file from the following package.
769067370 Dec 16 2021 BELD5_1US3_FIA6.1_2017CLD_Canada.nc4

There is another ncf (nc3) version of the 1km BELD5 file from the 2017platform.
33914800996 Jan 13 2021 BELD5_1US3_FIA6.1_2017CLD_Canada.ncf
I didn’t try the nc3 version of the 1km BELD5 file by thinking that the nc3 and nc4 versions have the same numbers.

Thanks for your help.


I think you are using the correct, most up to date BELD5 file. You are just comparing vs the 2017NEI b3grd which is older. Did you compare your b3grd file vs the 2016v2 b3grd file in this zipped file (


Not yet. Will compare and report back to here. Thanks!

I compared the 12US2 b3grd file we have generated to the windowed out 12US2 b3grd of 2016fj_16j from the 2016v2platform, and found similar differences. I further compared the original 12US1 b3grid3 from the 2017platform and the one from the 2016v2platform, and found they are identical. See the plot as follows:

Hi Jeff,

I compared the water percentage between the GRIDCRO2D file (from WRF3.9.1.1 run) and the BELD5 file (generated from the 1km BELD5 dataset) for both the 12US2 and GA4 grids. The BELD5 has a freshwater variable (MODIS_17) and an ocean water variable (MODIS_21), while GRIDCRO2D has only one water variable (LUFRAC_17). The differences are significant; please see the attached plots:

My understanding is that EPA recommends a water-fix to resolve such inconsistencies in water percentage between GRIDCRO2D and BELD4.1.

Does EPA recommend a similar water-fix resolution for the BELD5 data? If so, would you please share the fix?

Note that I have tried applying a similar water-fix for BELD4.1 to BELD5; but they differ from the EPA release b3grd file.




I think the only issue to resolve that you have seen with just the BELD5 data is the lakes in Canada and maybe a tiny amount along other coastlines correct? I will see what I can find out, but the impact on emissions is very small for these differences. EPA ORD moved to using MODIS for water in BELD5 and has continued to do so for BELD6 which is available. There is also BEIS4 now available in SMOKE v5. Are you interested in moving to BEIS4/BELD6 configuration? If not, I will see if I can find out where the BELD5 1km data is for the small water changes.


Hi Jeff,

Yes, the only issues are the differences along the coastlines and the fresh water body edges.

We are interested in BEIS4/BELD6 if they are available for user customized applications. But, for the current project, we have already used the 12km EPA b3grd file for our annual CMAQ modeling, we wanted to make our 4-km modeling consistent with the 12-km (using the same way to produce the b3grd file), if possible.

Thanks for your kindly help.